extend Return.convert with optional startType and periodicity functionality

[braverock]

Return.convert originally written by @erolbicero uses xts properties to detect the return calculation type of an input series and convert it to a different type.

Return.convert is also somewhat of an incomplete function and is not exported by PerformanceAnalytics, so we should make it more robust and then export it.

Not all time series of returns will contain the return type in the xts metadata. We should stop if there is no metadata and inform the user of the situation. In order to allow for the conversion of "unmarked" returns series by adding a startType field that takes the same arguments as destinationType which would define the return type we are converting "from".

Additionally, it would be nice for a "conversion" utility to have the ability to "convert" a time series to a lower periodicity, as it is often convenient to take e.g. intraday or daily data and convert it to lower periodicity (e.g. intraday to hourly or daily, daily to weekly or monthly, etc.)

Implementation notes (possibly non-exhaustive):

for the implementation of startType:

• add a default destinationType of 'NULL' to the existing list, in case the user just wants to lower the periodicity
• add stop/warning if the input series has no return type metadata, and suggest startType
• add startType argument to the function signature, with the same c(NULL, 'discrete','log','difference', 'level') options
• add stop/warn if startType is provided and the input series has a different return type in its metadata
• use the optional startType for conversion

for the implementation of periodicity:

• add dots to the function signature, and make sure they are passed appropriately to lower level functions as needed
• add period and k arguments to the function signature with a default period=NULL and k=1 (as with xts function to.period)
• check the periodicity of the input data
• stop/warn if the periodicity of the input data is higher than the periodicity requested in to_period
• if type conversion is also happening, do the type conversion before changing the periodicity
• use endpoints to determine the indexes of the k'th periods
• aggregate cumulative returns of the correct type by using period.apply over the endpoints
• handle multicolumn data which may have leading or trailing NA's

Additionally, we should create, document, and export a to.period.returns wrapper for Return.convert that aligns with the structure of xts's to.period to allow for easier user experience for users familiar with to.period. The wrapper should call Rturn.convert and provide additional arguments like name not included in Return.convert. (obviously the OHLC argument of to.period doesn't make sense for to.period.returns)

overall implementation requirements:

• add use of checkData to convert to xts internally
• add use of reclass to convert back to the input type if xts data was not provided
• fully document all the new options and behaviors
• add @examples to the documentation (and example snapshot update to the example tests)
• add a full set of testthat tests to the test suite, verifying both coverage and correctness of the type and periodicity conversions
• @export the functions